web-statistics as a craft

Some time ago, as a technical person, I couldn’t understand the complexity of the whole subject of web  statistics and analytics. 

  • Web-statistics – measuring how many pages were visited and when
  • Web-analytics –  trying to understand what it means

 When you know little, it’s always easy to build a logical solid theory about something.  That’s why people, who analyzed their first month of logs sound so sure about themselves. When you’re learning more, you find that your logical solid theory must be expanded. Then expanded some more and more. At some point you find out that your initial theory was completely inadequate. That’s a point on your learning curve called “I know that I don’t know anything”

I’ve reached that point about web-analytics on a training from Omniture. They have 2 days basic training on how to use their software. And 2 more days advanced training on that. And this is only for users, not for administrators!

OF course biggest part of the training was not statistics, but analytics, the way how to analyze your results. The subject is very complex. But actually, I figured out that simply the measurement of such objects, as “hit”, “visit”, “visitor”, “page view” is more a craft than a science. Read more here about key definitions in web-statistics.

So, why the web-statistics is complex:

  • The philosophy of underlying protocol  HTTP is not really helping the measurement. The protocol is session-less and asynchronous. You don’t exactly know what user has done on your website and weather it was successful, you have to guess
  • The philosophy of web browsing is not really designed for interactions. It assumes that user just reads text, clicks and reads further ahead. If user is trying to do anything else, like go back or reload or double the page, browsers react differently and sites react differently. Very hard to measure.
  • The philosophy of HTML is not really designed for modern interactive WWW. It has nothing to do directly with statistics, but it lead to creation of number of new technologies, like flash, java, AJAX etc. And those are often difficult to measure
  • What is a difference between pages anyway? Google Analytics defines it as change in the URL of your browser. This leads to a number of problems, like difference between / (just slash) and http://alleko.com/ (site name with slash). Also some technologies allow to change pages without changing URL
  • When user opens a heavy homepage of a site, his computer makes 50-100 or even more TCP connections to the web server. Sometimes is hard to define that it’s one hit.
  • It’s not obvious when to detect an end of a visit. User most often closes the browser when he doesn’t want to be on the site anymore. An industry standard says 30 minutes after the last hit


Different vendors have different approaches to the measuring, so often data from Google Analytics don’t match with Omniture or something else. Don’t worry, they’re not manufacturers, they’re crafters. Ironically there is no really standards in web-analytics either. Is 4% conversion high? Is 20% bounce rate low? Nobody could say for sure, every site is unique.

The only thing we can do is to compare comparable. Just choose one solution and measure your success against your success last month. Of year. You can not measure objectively.

You can leave a response, or trackback from your own site.

Sorry, no posts matched your criteria.